Why Social Data Could Power the Future of eCommerce Personalization November 30, 2011 by Michael Olson content personalization, customer profile data, ecommerce, social login Personalization is far from an exact science. Despite the torrent of consumer data and merchandising tools at an online retailer’s disposal, product recommendations on eCommerce sites today often are plagued by irrelevance, banality and inaccuracy. Coupled with privacy concerns, this has caused a segment of online shoppers to question the value of personalization. While consumers have been slow to cozy up to the idea, this isn’t because personalization is a poor merchandising strategy – retailers simply aren’t doing it right. So how do you recommend products that are actually aligned with a consumer’s current interests? Which data sources provide the best window into a shopper’s motivations? These are the questions eCommerce managers should be asking. The data within a consumer’s social network profile may hold the keys to effective personalization. But before explaining why, let’s start by looking at how eCommerce personalization works today. How Retailers Personalize Product Recommendations Today Online retailers traditionally have relied on one method to inform product recommendations – predictive modeling based on clickstream data and transaction data. Amazon.com pioneered the use of clickstream and transaction data for personalization years ago by recommending products based on your browsing behavior and purchase history. How does the technology actually work? Online retailers generally outsource product recommendations to specialized personalization engines. These vendors aggregate data from multiple sources, including browsing behavior and purchase history, and utilize modeling techniques to group people into clusters based on those attributes and behavior patterns. For example, consumers in San Francisco who viewed electronics products and purchased iPods online may also be interested in MacBook Pro laptops. These audience segments form the basis of behavioral targeting and personalization, but they are generated from archetypes or personas and not uniquely tied to real consumer identities. From a privacy perspective, this is a good thing. Unless you’ve opted in to sharing personal information with a retailer, do you really want that company or a data mining service tracking behaviors that are uniquely tied to your identity? The problem with the current methods of personalization is they often fall short at providing relevant and reliable recommendations for shoppers, and the data sources are not always stable or actionable for sustained use in marketing programs. Personalization engines dynamically serve content via client-side technology and widgets. These product recommendations are offered as part of the presentation layer on a site, but retailers often lack access to the underlying consumer data that informs a recommendation. And clickstream data is only as persistent as the tracking cookie placed in a user’s web browser upon her first visit to a site. Once that cookie is removed, a retailer relying on clickstream behavior to recommend products must start over from scratch. So, the clickstream is not dependable due to its fragility within a browser cookie. Nor is transaction data universally reliable as a predictor of intent, due to gift and experiential purchase behavior. That power tool purchase I made as a gift for Father’s Day does not indicate that I want to be served recommendations for similar products – power tools are not aligned with my interests. Similarly, recommendations for travel guides to Costa Rica long after my trip ended are not likely to induce future purchase intent. Here’s the other challenge with transaction data – purchase conversion rates for first-time eCommerce site visitors hover at about 3-4%, meaning that most of your visitors have never purchased a product from your site. Without a historical record of past purchases to inform recommendations, retailers are left in the dark. If clickstream data is unreliable and transaction data isn’t scalable across your audience or the best predictor of future purchases, then what is? A few retailers ahead of the curve in this space are hedging their bets on social data, specifically leveraging a consumer’s interests, real-time sentiments and social graph to inform recommendations. What is Social Data? People maintain a wealth of demographic and lifestyle information on their social networks, including gender, location, interests, likes, friends and a real-time stream of status updates. The building block to accessing this data with a consumer’s explicit permission is social login. When a shopper uses an existing account from social networks like Facebook, Twitter, Google, Yahoo or LinkedIn to register on your site, it results in a faster and easier registration process that eliminates the need to remember a site-specific password. For retailers and marketers, social login is a secure way for consumers to opt-in to sharing their social profile data with your site to expedite registration and checkout, and gain a more personalized shopping experience. Why Social Data Enables More Relevant Personalization I mentioned earlier that audience segments used in personalization algorithms are generated from archetypes and not uniquely tied to real consumer identities. But when a consumer explicitly shares her social profile data with you, the paradigm changes. Now you actually know the potential customer you’re targeting. Imagine if REI knew that I was really interested in bicycling and snow skiing from the moment I connected a social identity on its site, and could tailor product recommendations to prevent me from needing to wade through troves of camping or climbing gear. Social profile data makes this possible by providing retailers with opt-in access to a consumer’s interest graph – information such as likes, hobbies and real-time status updates, all of which can be appended to existing data structures to augment personalization efforts. But social data isn’t limited to your interests. Retailers such as Sears, Levi’s and Etsy are starting to leverage a consumer’s friends to create social shopping experiences. Sears.com lets customers share their list of friends from Facebook with the site, and then recommends gift ideas based on the birthdays and interests of those friends: Features debuted at Facebook’s recent f8 Conference will help further contextualize real-time status updates and enable more effective intent targeting. Beginning in 2010, Facebook deemphasized its generic share button in favor of the possibly just as generic like button. The problem retailers have faced with the like button (and the inherent difficulty with real-time targeting) is a lack of context. I just liked your product – does a retailer have any idea if I already own it or am now considering a purchase? The new sharing buttons introduced at f8 will enable consumers to share more contextual updates to their Facebook stream – instead of liking a product, I can specify that I want a product or just bought a product. For retailers using real-time data from social networks for personalization, this means better and more germane recommendations that are aligned with a shopper’s intent. In Tim Berners-Lee’s vision of the future web, technology will evolve to understand the meaning, or semantics, of information on the World Wide Web. This hasn’t happened yet, partly because computers are not smart enough but also because we’re not using the right data to predict or interpret consumer behavior. As the research will tell you, shoppers are still split on the value of personalization. The disparate sentiment is partially due to privacy concerns, but also an indictment of the effectiveness of current personalization methods. The next generation of eCommerce personalization will mitigate both by utilizing social data obtained with consumer permission to provide a more tailored, relevant shopping experience that genuinely influences the likelihood of purchase.